OpenAI Reportedly Bought Tens of Thousands of Macs for Reinforcement Learning

OpenAI has reportedly bought tens of thousands of Macs for AI training, and it is not buying laptops. According to The Information, OpenAI purchased large numbers of Mac mini and Mac Studio desktop machines specifically for reinforcement learning, buying out stock at one point and scrambling to secure more. Anthropic is doing similar work, renting Mac minis through Amazon Web Services. The machines are used to train "computer-use agents" — AI systems that operate a computer on their own to complete multi-step tasks such as editing and testing code, organizing email, and summarizing documents. The trend has already shown up in Apple's earnings: Mac sales grew nearly 29% year over year to $10.3 billion in the latest quarter, outpacing every other Apple product line including iPhone and iPad, making the Mac Apple's fastest-growing business. AI training has long been dominated by Nvidia GPUs, but Macs are being bought at scale in this reinforcement-learning niche because of unified memory. Apple's M-series chips use a single shared memory pool that both CPU and GPU access directly, avoiding the data-transfer bottleneck between VRAM and system memory in Nvidia GPUs. Unlike thin MacBooks, the Mac mini and Mac Studio ship with dedicated cooling systems, so they can run complex AI workloads for hours without thermal throttling — critical for reinforcement learning runs that last days. Apple is also promoting EXO Labs' open-source project that clusters multiple Macs to run trillion-parameter models locally, and the new Mac Studio emphasizes multi-machine clustering, with this product cycle arriving in August instead of the usual year-end refresh. The Mac's rise in local AI has caught Nvidia's attention. Sources say Nvidia now views Apple as its biggest competitor in local AI and launched DGX Spark, an AI desktop computer styled similarly to the Mac mini, late last year to target the same market. Apple, for its part, is struggling with supply: surging memory demand from AI data centers has caused an industry-wide shortage, and the high-end Mac mini and Mac Studio models most attractive to AI developers have been sold out for months. Former Apple AI marketing manager Todd Dailey says some enterprises have already turned to Nvidia's DGX Spark amid Mac supply constraints, and that the Mac's enterprise boom was an accident rather than a planned strategy. The opportunity has attracted new entrants: Peter Voell, a former OpenAI compute infrastructure employee, founded Mount Thor, a stealth cloud-computing company based on Apple hardware. Apple is betting on partners like Mount Thor and webAI to push Macs deeper into the enterprise market, and has begun building its own servers with Mac chips — though so far those servers are used only internally for Private Cloud Compute and are not sold.